Research On Video Moving Target Extraction And Tracking Algorithm

The purpose of Moving Target Detection and Tracking is extracted sequences ofMoving Targets from the continuous image, and tracked the Moving Target referred toin real-time. As important research techniques of Computer Image Processing andMachine Vision, it involves many areas such as Pattern Recognition、ArtificialIntelligence、Automatic Control. And on the field of Intelligent Control, RobotNavigation, Military Guidance, Intelligent Transportation, Medical Image Analysis, ithas a profound impact and a wide range of applications.Moving Objects Tracking are mainly divided into two aspects: The MovingTarget Detection and Moving Target Tracking. Moving Target Detection means todetect targets and extract the size, location and other related informations from thevideo stream in real-time. And Moving Target Tracking is to continue track anddetermine the trajectory from the Moving Target detected.Although there are many algorithms and methods for detection and follow-upstudy, no algorithm or method can effectively deals with any scene so far. SoDetection and Tracking becomes a hot issue of the scholar’s study and it also becomesa worldwide problem, such as: the Moving Target Block, Scaling, ComplexBackground and non-rigid body deformation. Detection and tracking that hasimportant practical and theoretical value for this reason.The thesis applies cluster thinking, forecasting judgment, Scale-space ideology,pattern matching and other commonly classical methods to the study of video MovingTarget Detection and Tracking Algorithm. And it tries to seek for a Moving TargetDetection and Tracking Algorithm that is real-time, robust and high-precision to solveabove problems effectively. The main contents include:1、In Moving Target Detection, this paper applies the Mean Shift Algorithm tothe movement of the Target Detection, in order to achieve the effective detection ofMoving Targets. Because it is parameter-free and fast iterative, and better robust tonoise, interference and other complex environment, the Mean Shift ClusteringAlgorithm has a very good effect on Image Segmentation and Image Smoothing. Butreal-time has to be considered when it will be applied to Moving Target Detection, sothis paper raises “Iterative Strategy” to solve this problem on the basis of thesimilarity of the Color Space, with improving the accuracy and real-time of theMoving Target Detection. And a simple three differential Edge Detection Algorithmhas been used to further improve the robustness of the Algorithm. 2、In Moving Target Tracking, this paper increases Scale-space Algorithms anddifferential method to track the moving targets based on the tracking frameworkwhich is combined with Kalman filtering prediction and existing Camshift Algorithm.And this paper takes advantage of the Kalman filter prediction algorithm to predict thetarget Moving position and related information in the next frame based on MovingTarget Detection Algorithm for Moving Target, and then matches optimization searchfor the results obtained and the target template fusion of the Scale-space CamshiftAlgorithm. And this paper corrects and updates target template according to theregional differential method, thereby tracks the Moving Targets accurately. In paper,the improved algorithm can effectively update the target template and amend windowand adapt to the rotation, scale change and deformation of the Moving TargetTracking with a higher accuracy than Traditional Algorithms.The algorithm mentioned is to be realized using Flash Builder4developmentplatform based on the Intel processor-based Windows environment in the paper.Seeing the effects obtained, the improved Moving Target Detection and TrackingAlgorithm mentioned in the paper is feasible, And it achieves accurate and robustextraction, and real-time tracking of the target to meet the requirements of real-timemonitoring, At the same time,the treatment effect of occlusion, deformation, scalesignificantly better than Traditional Tracking Algorithm.